A deep dilated convolutional residual network for predicting interchain contacts of protein homodimers

Deep learning has revolutionized protein tertiary structure prediction recently. The cutting-edge deep learning methods such as AlphaFold can predict high-accuracy tertiary structures for most individual protein chains. However, the accuracy of predicting quaternary structures of protein complexes consisting of multiple chains is still relatively low due to lack of advanced deep learning methods in the field. Because interchain residue-residue contacts can be used as distance restraints to guide quaternary structure modeling, here we develop a deep dilated convolutional residual network method (DRCon) to predict interchain residue-residue contacts in homodimers from residue-residue co-evolutionary signals derived from multiple sequence alignments of monomers, intrachain residue-residue contacts of monomers extracted from true/predicted tertiary structures or predicted by deep learning, and other sequence and structural features. Tested on three homodimer test datasets (Homo_std dataset, DeepHomo dataset, and CASP14-CAPRI dataset), the precision of DRCon for top L/5 interchain contact predictions (L: length of monomer in a homodimer) is 43.46%, 47.15%, and 24.81% respectively, which is substantially better than two existing deep learning interchain contact prediction methods. Moreover, our experiments demonstrate that using predicted tertiary structure or intrachain contacts of monomers in the unbound state as input, DRCon still performs reasonably well, even though its accuracy is lower than when true tertiary structures in the bound state are used as input. Finally, our case study shows that good interchain contact predictions can be used to build high-accuracy quaternary structure models of homodimers.

The pocketome of G protein-coupled receptors reveals previously untargeted allosteric sites

G protein-coupled receptors do not only feature the orthosteric pockets, where most endogenous agonists bind, but also a multitude of other allosteric pockets that have come into the focus as potential binding sites for synthetic modulators. We have investigated 557 GPCR structures to better characterise such pockets by exhaustively docking small molecular probes in silico and converting the ensemble of binding locations to pocket-defining volumes. Our analysis confirmed all previously identified pockets and revealed nine previously untargeted sites. In order to test for the feasibility of functional modulation of receptors through binding of a ligand to such sites, we mutated residues in two sites in two model receptors. Moreover, we analysed the correlation of inter-residue contacts with the activation states of receptors and showed that contact patterns closely correlated with activation indeed coincide with these sites.

Unique Features of Different Classes of G-Protein-Coupled Receptors Revealed from Sequence Coevolutionary and Structural Analysis

G-protein-coupled receptors (GPCRs) are the largest family of human membrane proteins and represent the primary targets of about one third of currently marketed drugs. Despite the critical importance, experimental structures have been determined for only a limited portion of GPCRs and functional mechanisms of GPCRs remain poorly understood. Here, we have constructed novel sequence coevolutionary models of the A and B classes of GPCRs and compared them with residue contact frequency maps generated with available experimental structures. Significant portions of structural residue contacts were successfully detected in the sequence-based covariational models. “Exception” residue contacts predicted from sequence coevolutionary models but not available structures added missing links that were important for GPCR activation and allosteric modulation. Moreover, we identified distinct residue contacts involving different sets of functional motifs for GPCR activation, such as the Na+ pocket, CWxP, DRY, PIF and NPxxY motifs in the class A and the HETx and PxxG motifs in the class B. Finally, we systematically uncovered critical residue contacts tuned by allosteric modulation in the two classes of GPCRs, including those from the activation motifs and particularly the extracellular and intracellular loops in class A GPCRs. These findings provide a promising framework for rational design of ligands to regulate GPCR activation and allosteric modulation.

Crystal structures of Val58Ile tryptophan repressor in a domain-swapped array in the presence and absence of L-tryptophan

The crystal structures of domain-swapped tryptophan repressor (TrpR) variant Val58Ile before and after soaking with the physiological ligand L-tryptophan (L-Trp) indicate that L-Trp occupies the same location in the domain-swapped form as in native dimeric TrpR and makes equivalent residue contacts. This result is unexpected because the ligand binding-site residues arise from three separate polypeptide chains in the domain-swapped form. This work represents the first published structure of a domain-swapped form of TrpR with L-Trp bound. The presented structures also show that the protein amino-terminus, whether or not it bears a disordered extension of about 20 residues, is accessible in the large solvent channels of the domain-swapped crystal form, as in the structures reported previously in this form for TrpR without N-terminal extensions. These findings inspire the exploration of L-Trp analogs and N-terminal modifications as labels to orient guest proteins that cannot otherwise be crystallized in the solvent channels of crystalline domain-swapped TrpR hosts for potential diffraction analysis.

DNCON2_Inter: predicting interchain contacts for homodimeric and homomultimeric protein complexes using multiple sequence alignments of monomers and deep learning

Deep learning methods that achieved great success in predicting intrachain residue-residue contacts have been applied to predict interchain contacts between proteins. However, these methods require multiple sequence alignments (MSAs) of a pair of interacting proteins (dimers) as input, which are often difficult to obtain because there are not many known protein complexes available to generate MSAs of sufficient depth for a pair of proteins. In recognizing that multiple sequence alignments of a monomer that forms homomultimers contain the co-evolutionary signals of both intrachain and interchain residue pairs in contact, we applied DNCON2 (a deep learning-based protein intrachain residue-residue contact predictor) to predict both intrachain and interchain contacts for homomultimers using multiple sequence alignment (MSA) and other co-evolutionary features of a single monomer followed by discrimination of interchain and intrachain contacts according to the tertiary structure of the monomer. We name this tool DNCON2_Inter. Allowing true-positive predictions within two residue shifts, the best average precision was obtained for the Top-L/10 predictions of 22.9% for homodimers and 17.0% for higher-order homomultimers. In some instances, especially where interchain contact densities are high, DNCON2_Inter predicted interchain contacts with 100% precision. We also developed Con_Complex, a complex structure reconstruction tool that uses predicted contacts to produce the structure of the complex. Using Con_Complex, we show that the predicted contacts can be used to accurately construct the structure of some complexes. Our experiment demonstrates that monomeric multiple sequence alignments can be used with deep learning to predict interchain contacts of homomeric proteins.

Unique Features of Different Classes of G-Protein-Coupled Receptors Revealed from Sequence Coevolutionary and Structural Analysis

G-protein-coupled receptors (GPCRs) are the largest family of human membrane proteins and serve as the primary targets of about one third of currently marketed drugs. Despite the critical importance, experimental structures have been determined for only a limited portion of GPCRs. Functional mechanisms of GPCRs remain poorly understood. Here, we have constructed sequence coevolutionary models of the A, B and C classes of GPCRs and compared them with residue contact frequency maps generated with available experimental structures. Significant portions of structural residue contacts have been successfully detected in the sequence-based covariational models. "Exception" residue contacts predicted from sequence coevolutionary models but not available structures added missing links that were important for GPCR activation and allosteric modulation. Our combined coevolutionary and structural analysis revealed unique features of the different classes of GPCRs. First, we provided evidence from coevolutionary couplings that dimerization is required for activation of class C GPCRs, but not for activation of class A and B GPCRs. Second, we identified distinct residue contacts involving different sets of functional motifs for activation of the class A and B GPCRs. Finally, we uncovered critical residue contacts tuned by allosteric modulation in the three classes of GPCRs. These findings provide a promising framework for designing selective therapeutics of GPCRs.

Recent Applications of Deep Learning Methods on Evolution- and Contact-Based Protein Structure Prediction

The new advances in deep learning methods have influenced many aspects of scientific research, including the study of the protein system. The prediction of proteins’ 3D structural components is now heavily dependent on machine learning techniques that interpret how protein sequences and their homology govern the inter-residue contacts and structural organization. Especially, methods employing deep neural networks have had a significant impact on recent CASP13 and CASP14 competition. Here, we explore the recent applications of deep learning methods in the protein structure prediction area. We also look at the potential opportunities for deep learning methods to identify unknown protein structures and functions to be discovered and help guide drug–target interactions. Although significant problems still need to be addressed, we expect these techniques in the near future to play crucial roles in protein structural bioinformatics as well as in drug discovery.

DNCON2_Inter: Predicting interchain contacts for homodimeric and homomultimeric protein complexes using multiple sequence alignments of monomers and deep learning

Deep learning methods that achieved great success in predicting intrachain residue-residue contacts have been applied to predict interchain contacts between proteins. However, these methods require multiple sequence alignments (MSAs) of a pair of interacting proteins (dimers) as input, which are often difficult to obtain because there are not many known protein complexes available to generate MSAs of sufficient depth for a pair of proteins. In recognizing that multiple sequence alignments of a monomer that forms homomultimers contain the co-evolutionary signals of both intrachain and interchain residue pairs in contact, we applied DNCON2 (a deep learning-based protein intrachain residue-residue contact predictor) to predict both intrachain and interchain contacts for homomultimers using multiple sequence alignment (MSA) and other co-evolutionary features of a single monomer followed by discrimination of interchain and intrachain contacts according to the tertiary structure of the monomer. We name this tool DNCON2_Inter. Allowing true-positive predictions within two residue shifts, the best average precision was obtained for the Top-L/10 predictions of 22.9% for homodimers, and 17.0% for higher order homomultimers. In some instances, especially where interchain contact densities are high, DNCON2_Inter predicted interchain contacts with 100% precision. We also developed Con_Complex, a complex structure reconstruction tool that uses predicted contacts to produce the structure of the complex. Using Con_Complex, we show that the predicted contacts can be used to accurately construct the structure of some complexes. Our experiment demonstrates that monomeric multiple sequence alignments can be used with deep learning to predict interchain contacts of homomeric proteins.

A Reproducibility Analysis-based Statistical Framework for Residue-Residue Evolutionary Coupling Detection

Direct coupling analysis (DCA) has been widely used to predict residue-residue contacts to assist protein/RNA structure and interaction prediction. However, effectively selecting residue pairs for contact prediction according to the result of DCA is a non-trivial task, since the number of highly predictive residue pairs and the coupling scores obtained from DCA are highly dependent on the number and the length of the homologous sequences forming the multiple sequence alignment, the detailed settings of the DCA algorithm, the functional characteristics of the macromolecule, etc. In this study, we present a general statistical framework for selecting predictive residue pairs through significant evolutionary coupling detection, referred to as IDR-DCA, which is based on reproducibility analysis of the coupling scores from replicated DCA. IDR-DCA was applied to select residue pairs for contact prediction for 150 proteins, 30 protein-protein interactions and 36 RNAs, in which we applied three widely used DCA software to perform the DCA. We show that with the application of IDR-DCA, the predictive residue pairs can be effectively selected through a universal threshold independent on the DCA software.

ConPlot: web-based application for the visualization of protein contact maps integrated with other data

Summary Covariance-based predictions of residue contacts and inter-residue distances are an increasingly popular data type in protein bioinformatics. Here we present ConPlot, a web-based application for convenient display and analysis of contact maps and distograms. Integration of predicted contact data with other predictions is often required to facilitate inference of structural features. ConPlot can therefore use the empty space near the contact map diagonal to display multiple coloured tracks representing other sequence-based predictions. Popular file formats are natively read and bespoke data can also be flexibly displayed. This novel visualization will enable easier interpretation of predicted contact maps. Availability and implementation available online at www.conplot.org, along with documentation and examples. Alternatively, ConPlot can be installed and used locally using the docker image from the project’s Docker Hub repository. ConPlot is licensed under the BSD 3-Clause. Supplementary information Supplementary data are available at Bioinformatics online.